by Jost Migenda, Talks Chair for RSECon26

RSECon is a conference organised by the community, for the community. From the annual open call to join the RSECon committee, to offering mentoring to less experienced presenters, to the Emerging Voices plenary, we make a deliberate effort to encourage new community members to participate—because we know: A Better Community = Better Software = Better Research.

Yet as both the community and RSECon have grown over the years, it is getting harder for many members to see what happens behind the scenes of the conference. In this blog post, we describe all four main stages of the selection process at the last several RSECons:

1. Submissions
2. Reviews
3. Ranking
4. Building the schedule

At the end of this blog post, we will also share a few early statistics from RSECon26 and explain why the selection process was especially difficult this year, with many well-reviewed submissions not making the final cut.

Submissions

The submission process is documented on the RSECon website. Anyone—independent of career stage, institutional association, etc.—can submit proposals via an online system, where they enter the title, abstract, author list, type of submission and other details.

In addition to talks and posters, RSECon offers several types of submissions that are less commonly found at traditional research conferences:

• Walkthroughs: similar to talks, but focusing on live demonstrations of a tool or workflow
• Workshops: hands-on sessions, where audience members can learn a new technology or contribute to a project
• Birds-of-a-Feather: community-building sessions that often combine lightning talks, panels or breakout groups

New at RSECon26, we introduced Research Software Papers in collaboration with the Journal of Open Source Software (JOSS). These papers undergo a more rigorous review process developed by JOSS in place of steps 2 and 3 of the regular RSECon selection process.

Reviews

In parallel with the submission period, RSECon has an open call for reviewers. Since we want the final programme to be of interest to the RSE community, we need a diverse group of reviewers that reflects the RSE community. People with different experience levels, different subject area backgrounds, from different countries, … are all welcome!

There is only one restriction: We want reviewers who are involved in the RSE community. (In 2025, some “immigration influencer” encouraged his followers to sign up as reviewers to RSECon and various other computer science conferences, claiming that this would help them apply for a U.S. green card. This led to an influx of reviewer applications with no connection to RSE whatsoever; as a result, we updated the reviewer form in 2026 to ask people to briefly tell us how they have been involved in the RSE community.)

After submissions close, there is a review period of two weeks. This year, we assigned an average of nine submissions to each reviewer, asking them to rate:

• The relevance to the RSECon audience
• The quality of the abstract
• The internal consistency of the submission (e.g., do content and outcomes match the requested length)
• Whether they recommend acceptance.

These responses are translated into a numerical score, which will be used to rank submissions. We also ask reviewers about the technical level of the submission and the expected audience size. These are not considered when ranking submissions, but will be used later when scheduling parallel sessions. Finally, we ask reviewers to provide constructive feedback to the author and, in a separate text field, additional remarks for the programme team.

Ranking

Once multiple independent reviews per submission are in, members of the programme team read all reviews. We pay special attention to submissions whose numerical ratings have a high standard deviation, indicating significant disagreement between reviewers. Normally, we give significant leeway to reviewers in these cases—different people bring different perspectives, and that is a strength of the RSE community! However, there are a few cases where one reviewer notes a serious flaw that others have missed, or where one reviewer has clearly misunderstood the submission or reviewing guidelines. These adjustments by the programme team affect fewer than 10% of submissions.

For the first time this year, a reviewer very obviously used an LLM for their reviews with no manual oversight. The LLM-written comments were often patronising and contained misleading or nonsensical advice; we have excluded those reviews from consideration. For future years, we have made a note to include a GenAI/LLM policy for authors and reviewers, making it clear what is and is not seen as acceptable use in the submission and review process.

After these steps, we sort submissions by the average of the adjusted reviewer ratings; the result is one ranked list per submission type.

While reading all reviews, we also check them for feedback that is inappropriate or poorly worded and requires edits before sharing it with submitters. Thankfully, this is extremely rare; at most, we see a few comments that sound a bit too harsh. (They probably aren’t even intended that way by the reviewers, but nuances of tone invariably get lost in written comments.)

Building the Schedule

Once these rankings are available, the programme chairs cluster submissions by topic across different submission types. At this stage, we typically decide to reject a small number of submissions—e.g. if the same person submits a talk and a walkthrough on closely related topics, we may only accept the highest-ranked one. We may also adjust timings—e.g., some workshops or birds-of-a-feather sessions may receive a shorter slot than requested.

The first rough draft of the schedule, produced at this point, combines the accepted submissions, invited keynotes, sponsor talks & workshops, the poster session, and more. It also still contains a few blocked slots, e.g. since some sponsors are not yet confirmed. The most highly ranked submissions we were unable to fit into the schedule are put on a short wait list to fill these slots (if they later become available) or fill in for any cancellations.

Finally, we share outcomes (accept/waitlist/reject) and reviewer feedback with submitters.

First Observations from RSECon26

RSECon set a new record this year, with 215 submissions across all types. The average quality of these submissions was incredibly high. For example, the median rating of all talks and walkthroughs was 16 (out of 19) points; out of 117 submissions in these two categories, over 100 were rated “accept” or “strong accept” by a majority of reviewers. Results were similar for other submission types. Unfortunately, the programme schedule and venue were able to accommodate less than half of them, leading to a record number of rejections as well.

While some conferences offer posters as a backup option to people whose talks were rejected, we are unable to do so at RSECon26. Due to space limitations of the venue, posters were as competitive as other tracks, with 25 acceptances from 58 submissions.

Back in February, before submissions even opened, we started discussions within the programme team about an experimental session of lightning talks at RSECon26. Considering the record number of submissions, this will also be a valuable opportunity to give more people a chance to present. Stay tuned for details in the coming weeks!

Final words

Many thanks to everyone who contributed so many amazing submissions & thoughtful reviews.
The RSE community is amazing! ❤️